Simplifying the Synthesis of Metal-Organic Frameworks

Conspectus:

Metal-organic frameworks (MOFs) are porous, crystalline materials constructed from organic linkers and inorganic nodes that have attracted widespread interest due to their permanent porosity and highly modular structures. However, the large volumes of organic solvents and additives, long reaction times, and specialized equipment typically required to synthesize MOFs hinder their widespread adoption in both academia and industry. Recently, our lab has developed several user-friendly methods for the gram-scale (1–100 g) preparation of MOFs. Herein, we summarize our progress in the development of high-concentration solvothermal, mechanochemical, and ionothermal syntheses of MOFs, as well as in minimizing the amount of modulators required to prepare highly crystalline Zr-MOFs. To begin, we detail our work elucidating key features of acid modulation in Zr-MOFs to improve upon current dilute solvothermal syntheses. Choosing an optimal modulator maximizes the crystallinity and porosity of Zr-MOFs while minimizing the quantity of modulator needed, reducing the waste associated with MOF synthesis. By evaluating a range of modulators, we identify the pK_a, size, and structural similarity of the modulator to the linker as controlling factors in modulating ability. In the following section, we describe two high-concentration solvothermal methods for the synthesis of Zr-MOFs and demonstrate their generality among a range of frameworks. We also target the M₂(dobdc) (M = Mg, Mn, Fe, Co, Ni, Cu, Zn, Cd; dobdc⁴⁻ = 2,5-dioxido-1,4-benzenedicarboxylate) family of MOFs for high-concentration synthesis and introduce a two-step preparation of several variants that proceeds through a novel kinetic phase. The high-concentration methods we discuss produce MOFs on multi-gram scale with comparable properties to those prepared under traditional dilute solvothermal conditions. Next, to further curtail solvent waste and accelerate reaction times, we discuss the mechanochemical preparation of M₂(dobdc) MOFs utilizing liquid amine additives in a planetary ball mill, which we also apply to the synthesis of two related salicylate frameworks. These samples exhibit comparable porosities to traditional dilute solvothermal samples but can be synthesized in just minutes, as opposed to days, and require under 1 mL of liquid additive to prepare ~0.5 g of material. In the following section, we discuss our efforts to avoid specialized equipment and eliminate solvent use entirely by employing ionothermal conditions to prepare a variety of azolate- and salicylate-based MOFs. Simply combining metal chloride (hydrate) salts with organic linkers at temperatures above the melting points of the salts affords high-quality framework materials. Further, ionothermal conditions enable the syntheses of two new Fe(III) M₂(dobdc) derivatives that cannot be synthesized under normal solvothermal conditions. Last, as a demonstrative example, we discuss our efforts to synthesize 100 g of high-quality Mg₂(dobdc) in a single batch using a high-concentration (1.0 M) hydrothermal synthesis. Our Account will be of significant interest to researchers aiming to prepare gram-scale quantities of MOFs for further study.

GRAPHICAL ABSTRACT

1. Introduction.

Metal-organic frameworks (MOFs) are porous, crystalline coordination polymers constructed from organic linkers and inorganic secondary building units (SBUs) (Figure 1).¹ These materials have found widespread potential applications in gas storage and separation, catalysis, drug delivery, and beyond.^2–4 Due to the origins of the field in crystal engineering, MOFs are typically prepared under highly dilute solvothermal conditions in amide solvents such as N,N-dimethylformamide (DMF) with long reaction times (days to weeks).⁵ Dilute solvothermal syntheses are prevalent because the kinetics of the reaction between linkers and metal centers are easily biased by changes in concentration; dilute conditions discourage their reaction and enhance the reversible self-assembly of MOFs by allowing more opportunities for improper linkages to be corrected. However, technoeconomic analyses suggest that the large volumes of solvents used to prepare MOFs contribute greatly to their costs on scale.⁶ Large excesses (1–500 equiv. relative to linker) of additives that mediate crystal nucleation and growth processes, termed “modulators,” are often employed during MOF synthesis as well, leading to significant additional waste.⁷

Figure 1.

a) Structure of M₂Cl₂(btdd) (M = V, Mn, Fe, Co, Ni, Cu, btdd²⁻ = bis(1,2,3-triazolo[4,5-b],[4′,5′-i])dibenzo[1,4]dioxin), the secondary building unit, and the organic linker. Black, gray, white, blue, red and green spheres correspond to nickel, carbon, hydrogen, nitrogen, oxygen, and chlorine, respectively. b) Representative dilute synthesis of M₂Cl₂(btdd) (M = V, Mn, Fe, Co, Ni, Cu). c) Synthesis of Cu₂Cl₂(btdd) on 1.0 g of linker scale.

A related but less-appreciated challenge is the gram-scale synthesis of MOFs in the laboratory, which can be difficult because of the liters of solvent needed to make even one gram of MOF. For example, following the literature synthesis, the framework Co₂Cl₂(btdd) (btdd²⁻ = bis(1,2,3-triazolo[4,5-b],[4′,5′-i])dibenzo[1,4]dioxin) (Figure 1a) requires 1.5 L of organic solvent and 10 days to synthesize 1.0 g of MOF, with additional solvent and time needed to wash and purify the material (Figure 1b,c.).⁸ As such, there is a pressing need to develop new, reliable, and reproducible methods amenable to the preparation of MOFs on large scale. Equally pertinent is the development of simple syntheses to afford non-experts the opportunity to study MOFs for novel applications, such as in organic synthesis.

As MOFs transition from academic curiosities to industrially relevant materials, their scalable synthesis has emerged as a major area of study.⁹ Several sustainable and scalable alternatives to traditional solvothermal MOF syntheses have been developed.^5,10,11 User-friendly approaches to reduce the volume of organic solvents required for MOF synthesis include hydrothermal,¹² high-concentration solvothermal,¹³ ionothermal,¹⁴ and continuous flow methods.¹⁵ Using more specialized equipment (e.g., ball mills or extruders), MOFs can also be synthesized under mechanochemical conditions with minimal solvent use while cutting synthesis times down to the order of minutes.¹⁶

Herein, we detail our efforts to streamline the gram-scale syntheses of MOFs by reducing barriers to their scalable preparation. We begin with our work to identify superior modulators employed in traditional dilute solvothermal syntheses (Section 2). We then discuss representative work from our lab utilizing high-concentration solvothermal (Section 3), mechanochemical (Section 4), and ionothermal (Section 5) methods to greatly reduce the volumes of organic solvent required for MOF synthesis. We conclude with an illustrative example of our efforts to scale up the MOF Mg₂(dobdc) (dobdc⁴⁻ = 2,5-dioxidobenzene-1,4-dicarboxylate) (Section 6) and provide future directions for the field (Section 7). Overall, we hope this account will prove useful to those aiming to produce a known or new MOF on multi-gram scale for further study.

2. Improving acid modulation.

Depending on the strength of the metal–linker bonds, some MOFs require strict control over the synthesis conditions to facilitate reversible self-assembly, often making it challenging to produce high-quality samples without employing dilute solvothermal conditions. For example, Zr-based MOFs prepared under solvothermal conditions tend to exhibit superior crystallinity and porosity compared to samples prepared under alternative, solvent-free conditions.¹⁷ This is due, in part, to the strong Zr–O bonds comprising these frameworks, which limit the reversibility of self-assembly. On one hand, the robustness of Zr-MOFs has enabled their use in drug delivery and catalysis,^18,19 as well as in emerging applications such as water harvesting.²⁰ However, to synthesize Zr-MOFs of suitable quality for these applications, synthetic additives that suppress the rapid formation of poorly crystalline products are often required. The addition of monodentate acids—referred to as acid modulation—enhances the reversibility of Zr-MOF self-assembly by 1) protonating linkers bound to the nodes and/or 2) competing with the linkers (as their conjugate bases) for coordination to the nodes.⁷ The former mechanism, exogenous acid modulation, is favored using strong acids such as trifluoroacetic acid and hydrochloric acid. The latter pathway, competitive coordination, is best accomplished using monotopic carboxylic acids with conjugate bases of comparable nucleophilicity to linker molecules, such as benzoic acid/benzoate.

Although acid modulation is an effective and widely used approach to improve the crystallinity of Zr-MOFs, it is also highly wasteful. A substantial excess (1–500 equiv. relative to the linker) of modulator is often included to maximize the crystallite size or other material properties of Zr-MOFs; the modulator is then generally discarded with the supernatant after a single use.⁷ For example, a recently reported Zr-MOF from our lab, CORN-MOF-5 (CORN = Cornell University), requires 200 equiv. of formic acid during its synthesis to form high-quality, phasepure material.²¹ Given the breadth of Zr-MOFs that have been synthesized,¹⁸ many modulators have been tested that differ significantly in pK_a and electronic and steric effects.⁷ Few guiding principles exist for choosing the optimal modulator for a given Zr-MOF, contributing to the tendency among practitioners to simply add more of an unoptimized acid modulator to improve a material’s quality rather than identify a better modulator that can be employed in smaller amounts.

Using the Zr-MOFs UiO-66 and its isoreticular, pore-expanded analog UiO-68-Me₂ (UiO = Universitetet i Oslo) as representative materials,²² we systematically evaluated over twenty acid modulators to uncover the key principles that govern the effectiveness of acid modulation (Figure 2).²³ To the standard UiO-66 solvothermal synthesis conditions, we added 10 and 50 equiv. of acid modulators ranging in pK_a from −2 to 10. This includes inorganic acids, sulfonic acids, and carboxylic acids with divergent steric and electronic effects (Figure 2a). The efficacy of each modulator was principally evaluated using the volume-weighted average crystalline domain sizes (LVol-IB) of the corresponding UiO-66 and UiO-68-Me₂ samples, as quantified by whole powder pattern fitting of their experimental powder X-ray diffraction (PXRD) patterns.²⁴ These values are determined by fitting the whole powder pattern and take factors such as crystal strain into account, so they are more reliable than average crystallite sizes calculated through traditional Scherrer analysis. These crystalline domain sizes were substantiated by measuring the sizes of apparent crystallites in scanning electron microscopy (SEM) images of each sample.

Figure 2.

a) Acid-modulated synthesis of UiO-66. b) Volume-weighted average crystalline domain sizes (LVol-IB) vs. pK_a of UiO-66 samples prepared using 10 or 50 equiv. of acid modulators. MsOH = methanesulfonic acid, HCl = hydrochloric acid, TFA = trifluoroacetic acid, 2-TP = 2-thiophenecarboxylic acid, FA = formic acid, 3-TP = 3-thiophenecarboxylic acid, BA = benzoic acid, AA = acetic acid, PA = pivalic acid, PhOH = phenol. The gray box indicates the range between the pK_a1 and pK_a2 values of terephthalic acid. ^★ indicates modulators for which impurities were observed at 50 equiv. No MOF was formed when 50 equiv. of MsOH was used. Adapted with permission from ref.²³ Copyright 2022 American Society of Chemistry.

The effect of modulator pK_a on calculated LVol-IB values was substantial and differed depending on the concentration of the modulator. At lower concentrations (10 equiv.), the average crystalline domain sizes of UiO-66 samples steadily increase as the pK_a lowers below approximately 3.5 (Figure 2b). Methanesulfonic acid (MsOH), the strongest acid evaluated that does not completely inhibit MOF formation, produces the most crystalline sample of UiO-66 at 10 equiv., with a calculated LVol-IB value of nearly 400 nm. At this concentration, exogenous acid modulation by strong acids is more productive than competitive coordination. However, we observed a secondary increase in average crystalline domain size in the pK_a range of approximately 3.5–4.8, matching the range between pK_a1 and pK_a2 of the linker terephthalic acid (H₂bdc) (3.51 and 4.82, respectively), which we attribute to modulation via competitive coordination. This mechanism of modulation is dominant when the nucleophilicities of the linker and modulator conjugate base are similar.

The magnitude of the spike in LVol-IB values around the pK_a1 and pK_a2 of H₂bdc substantially increased upon raising the modulator concentration. At 50 equiv. of modulator, strong acids lead to impurity phases. Modulators in the pK_a range of the linker are more productive overall. In particular, benzoic acid formed among the largest crystalline domains of UiO-66 observed in our work (LVol-IB = 740 ± 40 nm), which we attribute to its structural similarity to the bdc²⁻ linker. Specifically, the aromatic backbone of benzoic acid likely enables it to engage in similar intermolecular interactions as the linker (e.g., π–π interactions) during MOF formation. As such, it mimics the linker more effectively than other modulators and is a stronger competitor for coordination to the Zr nodes.

To gain further insight into the efficacy of benzoic acid as a modulator, we evaluated the performance of a range of ortho- and para-substituted benzoic acid derivatives, including those bearing substituents of varying sizes and electronic effects.²³ PXRD and SEM analysis of the resultant MOFs revealed a clear trend in crystalline domain size related to the size of the substituent on the modulator: smaller acids such as benzoic acid and 4-fluorobenzoic acid are much more capable modulators than bulkier acids such as 4-tert-butylbenzoic acid. The presence of a linear free energy relationship between substituent size (measured through Charton or Sterimol parameters) and LVol-IB values further supports the inverse relationship between modulator size and modulating ability. Attempts to relate crystalline domain sizes of samples to electronic effects through Hammett σ values revealed no clear trend between modulating ability and the electronic nature of para-substituted benzoic acids. Together, our findings support two major criteria for effective coordination modulation: 1) structural similarity of the modulator to the linker and 2) small size to enable modulator diffusion through the forming framework. Applying these criteria, we hypothesized that five-membered aromatic heterocyclic carboxylic acids—specifically, 2- and 3-thiophenecarboxylic acids (2-TP and 3-TP, respectively)—should outperform benzoic acid. Consistently, under identical reaction conditions, both 2-TP and 3-TP lead to larger crystalline domains of UiO-66 at both low (10 equiv.) and high (50 equiv.) concentrations, making them the most productive modulators identified in our study (Figure 2b). When added to the synthesis of UiO-68-Me₂, 2-TP and 3-TP lead to large crystalline domains as well, suggesting that the criteria outlined above may be general among Zr-MOFs. By applying these findings, researchers will be able to create superior-quality MOFs while reducing the waste associated with their synthesis, facilitating the gram-scale solvothermal synthesis of highly crystalline Zr-MOFs.

3. High-concentration solvothermal synthesis.

Although better modulators help to decrease the waste associated with Zr-MOF synthesis, they do not address the primary barrier to producing MOFs on scale: the reliance on highly dilute (≤0.01 M in linker) solvothermal conditions. The simplest solution to this challenge would be to perform MOF syntheses at much higher reaction concentrations (e.g., 1.0 M in linker). However, examples of successful MOF syntheses even at intermediate concentrations (> 0.25 M in linker) are rare.^25–27 Instead, poorly crystalline or low surface area materials are typically obtained due to the rapid precipitation of frameworks from solution.^25,26,28 Our poor understanding of whether MOF self-assembly is possible at high concentrations motivated us to systematically study the high-concentration synthesis of ZrMOFs as a model system (Figure 3).¹³

Figure 3.

a) Synthesis of UiO-66 from either ZrCl₄ or ZrPiv. b) PXRD patterns of UiO-66 samples prepared using ZrCl₄ or ZrPiv and a [H₂bdc] of either 0.01 or 1.0 M. The simulated patterns based on the single-crystal X-ray diffraction (SCXRD) structures of ZrPiv and UiO-66 are included for reference. Ordered defect domains with a reo topology are indicated (*).³⁰ c) SCXRD structure of ZrPiv and UiO-66. The gray, red, and light blue spheres represent carbon, oxygen, and zirconium, respectively. Hydrogens are omitted for clarity. d) Left: UiO-66–1.0 M (ZrCl₄) (top) and UiO-66–1.0 M (ZrPiv) (bottom). Right: UiO-66–1.0 M (ZrCl₄) (left) sinking in water, in contrast to UiO-66–1.0 M (ZrPiv) (right) floating on water. Adapted with permission from ref.¹³ Copyright 2023 American Society of Chemistry.

Typical solvothermal Zr-MOF syntheses utilize Zr salt precursors (e.g., ZrCl₄) that first self-assemble into Zr₆ oxo clusters in the presence of water, which constitute the nodes of the frameworks. However, previous mechanochemical syntheses of Zr-MOFs generally use pre-formed Zr node precursors, indicating that node self-assembly may be problematic in the absence of large volumes of solvent.²⁹ To investigate whether node self-assembly is a limiting factor under high-concentration conditions, we sought a Zr₆ cluster precursor amenable to high-concentration solvothermal synthesis. We identified the pivalate-capped cluster Zr₆O₄(OH)₄(OPiv)₁₂ (OPiv⁻ = pivalate), referred to herein as ZrPiv (Figure 3c), as a potential yet previously unexplored MOF precursor. We developed a high-concentration (0.5 M) solvothermal synthesis of ZrPiv that avoids the use of water-sensitive Zr-alkoxide precursors and yields over 12 g of material in a single batch. This procedure could be extended to produce at least 6 g of the previously unreported Hf analogue, Hf₆O₄(OH)₄(OPiv)₁₂ (HfPiv), as well.²²

We next evaluated concentrations at which the archetypal Zr-MOF UiO-66 can be prepared under solvothermal conditions (Figure 3a). To further minimize the waste associated with MOF synthesis, we employed stoichiometric linker:metal ratios and avoided the use of exogenous acid modulators where possible, such as none for UiO-topology frameworks. Using either ZrCl₄ or ZrPiv as the node precursor at concentrations up to 1.0 M in DMF with stirring yields crystalline UiO-66 that is comparable to material prepared under the standard dilute conditions (Figure 3b). In particular, these high-concentration samples exhibit high Brunauer-Emmett-Teller (BET) surface areas of 1327 ± 8 and 1232 ± 7 m²/g for UiO-66–1.0 M (ZrCl₄) and UiO-66–1.0 M (ZrPiv), respectively.¹³

Although both ZrCl₄ and ZrPiv yield high-quality MOF, we observed some key differences between the two precursors. First, the PXRD of UiO-66–1.0M (ZrCl₄) (Figure 3b) contains additional low-angle reflections that correspond to crystalline domains of ordered missing node defects.³⁰ In contrast, UiO-66–1.0M (ZrPiv) does not display these reflections despite containing a similar number of overall defects, as determined by pore-size distribution and pore volume analysis. We hypothesize that this is due to the large amounts of HCl generated during high-concentration syntheses with ZrCl₄, as HCl is known to promote defect formation in Zr-MOFs.³¹ Second, samples prepared from ZrPiv require much longer reaction times (72 h) compared to only 1 h when using ZrCl₄, which is likely due to the slow exchange of the capping pivalates for linkers. Lastly, the use of ZrPiv leads to the incorporation of capping pivalates as missing-linker defects, producing MOF with a super-hydrophobic surface compared to the hydrophilic surface of UiO-66 prepared from ZrCl₄ (Figure 3d). Depending on which factor is more important for a given application—expediency or the incorporation of property-altering defects—ZrCl₄ or ZrPiv will be the optimal precursor for MOF synthesis.

Having synthesized high-quality UiO-66 at concentrations up to 1.0 M, we then evaluated the generality of high-concentration solvothermal synthesis for Zr- and Hf-MOFs. Using both HfCl₄ and HfPiv, crystalline and porous UiO-66(Hf) can be prepared at a concentration of 1.0 M in DMF. For substituted UiO-66 variants, UiO-66-NH₂ and UiO-66-(OH)₂, and expanded pore frameworks, UiO-67 and UiO-68-Me₂, we synthesized highly porous MOFs at concentrations up to 0.5 M using ZrPiv and up to 1.0 M using ZrCl₄ (Table 1).

Table 1.

77 K N₂ BET surface areas of MOFs prepared at high concentrations (HC) using either metal chloride salt or MPiv (M = Zr or Hf) precursors. Literature values reported for dilute solvothermal equivalents are included for reference.¹³

MOF	BET Surface Area (m2/g)
	Metal Salt (HC)	MPiv (HC)	Reported
UiO-66	1327	1232	1170
UiO-66(Hf)	961	988	950
UiO-66-NH2	822	765	815
UiO-66-(OH)2	464	503	560
UiO-67	2492	1954	2250
UiO-68-Me2	1237	2413	2470
MOF-808	1789	1732	2060
PCN-128	1321	1983	2384

We next extended our findings to the synthesis of Zr-MOFs with higher topicity linkers. The addition of carboxylic acid modulators was required to cap the nodes in these MOFs; however, their use was kept to a minimum by maintaining the modulator to solvent volume ratio used in reported dilute syntheses. Crystalline and porous MOF-808, a Zr-MOF constructed from tritopic linkers and six-connected nodes, was synthesized at concentrations up to 0.5 M in DMF with formic acid (53 equiv.) as a modulator using either ZrOCl₂‧8H₂O (standard for MOF-808 syntheses)²⁷ or ZrPiv as a precursor. Similarly, crystalline PCN-128, a Zr-MOF with eight-connected nodes and four-connected linkers that features 44 Å channels, could be synthesized at a linker concentration of 0.25 M in DMF using trifluoroacetic acid (5 equiv.) as a modulator and either ZrCl₄ or ZrPiv as a precursor. Notably, nearly all high-concentration samples prepared exhibit BET surface areas comparable to MOFs prepared under dilute conditions (Table 1). Two exceptions, UiO-68-Me₂ and PCN-128 prepared from ZrCl₄, exhibit surface areas that are lower than expected, especially compared to the surface areas of the corresponding MOFs prepared from ZrPiv. These two MOFs contain the largest pore apertures among the evaluated frameworks, indicating that self-assembly of the nodes may limit the high-concentration synthesis of large-pore MOFs. Nonetheless, this challenge can be overcome using ZrPiv as a pre-assembled node precursor for large-pore MOF synthesis at high concentrations.

Encouraged by our success, we set out to explore the high-concentration synthesis of other frameworks. One target family of MOFs is the MOF-74, CPO-27, or M₂(dobdc) (M = Mg, Mn, Fe, Co, Ni, Cu, Zn, Cd; dobdc⁴⁻ = 2,5-dioxido-1,4-benzenedicarboxylate) series.³² These frameworks are promising for gas storage and separations but are often difficult to produce on gram scale, as they are typically synthesized under dilute (~0.01 M) conditions in DMF/alcohol mixtures. We began by evaluating the synthesis of Mg₂(dobdc) at linker concentrations varying from 0.01 M to 1.50 M with stirring.³² PXRD analysis of the resulting materials confirmed that phase-pure Mg₂(dobdc) was obtained at concentrations up to 0.05 M, but at 0.08 M, the MOF was contaminated with a previously unreported crystalline phase that we term CORN-MOF-1 (Mg). By further increasing the concentration (0.10–1.50 M), phase-pure CORN-MOF-1 (Mg) can be obtained.

The structure of CORN-MOF-1 (Mg) was elucidated using a variety of spectroscopic, crystallographic, and computational techniques. CORN-MOF-1 is composed of H₂dobdc²⁻ linkers with protonated phenol groups and charge-balancing formate groups (generated via DMF hydrolysis) for an overall proposed molecular formula of Mg₂(H₂dobdc)(HCO₂)₂. CORN-MOF-1 (Mg) is likely favored at high reaction concentrations because of the reduced concentration of dimethylamine generated via the hydrolysis of DMF, which leads to only partial deprotonation of H₄dobdc to H₂dobdc²⁻ (and not dobdc^4–).²⁸ Notably, this high-concentration synthesis yields > 5 g of crystalline and porous CORN-MOF-1 (Mg) (BET surface area of 403 m²/g) in a single batch using only 10 mL of solvent.

We hypothesized that it should be possible to fully convert CORN-MOF-1 (Mg) into Mg₂(dobdc) upon further treatment with base, offering a simple route to prepare Mg₂(dobdc) on gram scale. Indeed, after extended heating (5 days) in DMF, a suspension of CORN-MOF-1 (Mg) was fully converted into phase-pure Mg₂(dobdc) via a solid-to-solid transformation, as confirmed by PXRD and SEM. This two-step, high-concentration process enables the preparation of > 4 g of crystalline and highly porous Mg₂(dobdc) in a single batch. This method was also used to prepare other variants of M₂(dobdc) in gram-scale quantities. Specifically, the Ni and Mn variants of CORN-MOF-1 were synthesized at reaction concentrations of 1.00 M and then converted to Ni₂(dobdc) and Mn₂(dobdc) samples with high BET surface areas.

An interesting question is why Zr-MOFs tend to form the same phase at low and high linker concentrations,^13,25 whereas M₂(dobdc) materials undergo a phase transition to a new kinetic product at high reaction concentrations.^28,32 We propose that Zr- and Hf-MOFs are especially well-suited for high-concentration synthesis because their M₆ nodes are robust and well-defined and thus strongly direct linker coordination into specific structures, even at high concentrations. In contrast, the metal centers that form the backbone of M₂(dobdc) materials can engage in many different coordination environments with the linkers during MOF self-assembly. As such, unintended topologies may be isolated under non-ideal conditions. However, our findings demonstrate that such kinetic phases may still serve as scalable intermediates to the desired frameworks.³² Overall, high-concentration synthesis represents a straightforward method to synthesize frameworks on multi-gram scale with reduced solvent use.

4. Mechanochemical synthesis.

To reduce solvent waste even further, and in some cases eliminate solvent use completely, researchers have turned to the mechanochemical synthesis of MOFs. Mechanochemistry, the initiation of a chemical reaction via a mechanical process such as grinding or extrusion, is a well-established synthetic technique for preparing polymers, alloys, battery materials, and pharmaceuticals.¹⁶ Among mechanochemical techniques, a number of successful methods have been employed to synthesize porous materials in minutes instead of the hours or days required for traditional solvothermal methods. In neat grinding (NG), reagents are combined in a ball mill, planetary mill, or twin-screw extruder without any exogenous liquids. Typically reliant on metal oxide or acetate salts as basic precursors, this process has been shown to enable the preparation of simple MOFs such as zeolitic imidazolate frameworks (ZIFs) and is scalable via twin screw extrusion.³³ The addition of liquids into the milling process, known as liquid assisted grinding (LAG), improves the crystallinity of MOFs compared to NG and enables access to MOFs that cannot be prepared via NG, such as MOF-801.³⁴ In good agreement with the findings of our laboratory (Section 2), cluster precursor synthesis has been found to assist mechanochemical MOF formation in some cases.³⁵ Mechanochemically synthesized MOFs are typically less crystalline than the corresponding MOFs prepared under solvothermal conditions, illustrating the challenge of preparing MOFs under such harsh physical conditions.

Mechanochemical syntheses of M₂(dobdc) MOFs remain underutilized compared to simple carboxylate MOFs and ZIFs. Previously reported mechanochemical syntheses of M₂(dobdc) analogs have primarily involved LAG of a basic metal oxide or acetate salt and the organic linker H₄dobdc in the presence of a small amount of solvent such as DMF, water, or methanol.³⁶ Among these frameworks, only the mechanochemical synthesis of Zn₂(dobdc) has been studied in detail.³⁷ In traditional solvothermal syntheses, the role of the amide solvent is to decompose at high temperatures (>100 °C) to produce amines (e.g. N,N-dimethylamine) that deprotonate the conjugate acid of the linker prior to MOF formation. Thus, we envisioned that disentangling the metal salt and base via the use of an exogenous liquid base could enable access to higher-quality materials by improving their reversible self-assembly. In particular, the use of a liquid amine as both the base and additive to assist LAG should enable the mechanochemical synthesis of these frameworks from simple metal salt precursors while bypassing the use of DMF. While the use of exogenous base to promote the solvothermal synthesis of M₂(dobdc) analogs is precedented,³⁸ its use under mechanochemical conditions had not been reported prior to our work (Figure 4).³⁹ We initially chose Hünig’s base (N,N-diisopropylethylamine) as the organic base for LAG, as its basicity (pKb = 3.2) is similar to triethylamine (pKb = 3.2) and methylamine (pKb = 3.3), which have previously been used to prepare M₂(dobdc) analogs under solvothermal conditions.⁴⁰

Figure 4.

a) Synthesis of M₂(dobdc) and Mg₂(dobpdc) under mechanochemical conditions. b) Synthesis of Mg₂(m-dobdc) under solvothermal (left) and mechanochemical (right) conditions. c) PXRD patterns of Mg₂(m-dobdc) under mechanochemical conditions (blue) and solvothermal conditions (red). The simulated pattern based on the SCXRD structure of the isostructural Co₂(m-dobdc) is included for reference. d) 30 °C CO₂ and adsorption isotherms in Mg₂(m-dobdc)-ST (red) and Mg₂(m-dobdc)-MC (blue). The lines correspond to individual fits to the dual-site Langmuir model. A data point was considered equilibrated when < 0.01% pressure change occurred over a 30 s interval. Adapted with permissions from ref.^{19, 42} Copyright 2020, 2022 Royal Society of Chemistry.

Beginning with Mg₂(dobdc), we found that grinding together Mg(NO₃)₂‧6H₂O and H₄dobdc with the addition of 4.4 equiv. of Hünig’s base at room temperature for five minutes in a planetary ball mill leads to high-quality MOF, as confirmed by PXRD and surface area analysis (Figure 4a,c, Table 2). Notably, this method can be effectively applied to the synthesis of numerous other M₂(dobdc) analogs (M = Mn, Co, Ni, Cu, and Zn). This approach is also amenable to the synthesis of other salicylate-based framework materials. Specifically, the isomeric Mg₂(m-dobdc) (m-dobdc⁴⁻ = 4,6-dioxido-1,3-benzenedicarboxylate) and the isoreticular, pore-expanded Mg₂(dobpdc) (dobpdc⁴⁻ = 4,4′-dioxido-3,3′-biphenyldicarboxylate) frameworks can be synthesized using the same approach, representing the first mechanochemical syntheses of these MOFs (Figure 4a,b). Aside from Mn₂(dobdc), which in our experience is unstable to air exposure, all of the prepared frameworks demonstrate Langmuir surface areas that are comparable to reported values for traditionally prepared samples. In fact, the mechanochemical synthesis of Mg₂(m-dobdc) was the first reported route to prepare permanently porous samples of this framework, albeit with modest crystallinity; it is non-porous when prepared using the originally reported solvothermal route.⁴¹ Subsequently, we reported a modified solvothermal procedure under dilute conditions (0.03 M in linker) that yields porous material with slightly improved crystallinity relative to the mechanochemically synthesized material (Figure 4c).⁴²

Table 2.

77 K N₂ Langmuir surface areas of MOFs prepared mechanochemically with liquid amine base. Experimental or literature values reported for solvothermal equivalents are included for reference.

MOF	Langmuir Surface Area (m2/g)
	Mechanochemical	Solvothermal
Mg2(dobdc)	1992	1914
Mn2(dobdc)	385	1797
Co2(dobdc)	1334	1438
Ni2(dobdc)	1281	1574
Cu2(dobdc)	1115	1515
Zn2(dobdc)	1204	1277
Mg2(m-dobdc)-Hünig’s	1793	1971
Mg2(m-dobdc)-Et3N	1964
Mg2(dobpdc)	3137	3780

The potential utility of Mg₂(m-dobdc) for gas storage and separations motivated us to further improve upon its mechanochemical synthesis. Simply switching the base from Hünig’s base to triethylamine provides a reproducible method for preparing high-quality Mg₂(m-dobdc) on 0.5 g scale. Mechanochemically synthesized Mg₂(m-dobdc) exhibits a BET surface area of 1653 m²/g, which exceeds that of its solvothermal equivalent (1556 m²/g). Furthermore, Mg₂(m-dobdc) prepared under mechanochemical conditions possesses one of the highest CO₂ capacities reported to date (6.14 mmol/g) for a porous material under simulated coal flue gas conditions (150 mbar, 40 °C) (Figure 4d), representing a key advance in post-combustion carbon capture. Currently, the use of exogenous base for LAG remains limited to salicylate frameworks. Despite this limitation, our findings support that, when combined with synthetic rationale and design, mechanochemical synthesis can assist in the rapid preparation of MOFs that are of higher quality than their solvothermal counterparts in some cases.

5. Ionothermal synthesis.

Another approach to preparing MOF that minimizes the use of organic solvent yet avoids the harsh grinding of mechanochemistry is ionothermal synthesis, in which an ionic liquid or molten metal salt is used as a solvent or structure directing agent.⁴³ Critically, this method also eliminates the need for specialized equipment such as ball mills, making access to complex frameworks remarkably simple and cost-effective. Ionothermal methods are well established for the preparation of covalent triazene frameworks (CTFs), in which a molten metal salt acts as a Lewis acid catalyst to drive triazene formation from nitrile-functionalized monomers.⁴⁴ Previously, our lab has reported the novel synthesis of a series of conjugated microporous polymers (CMPs) and porous aromatic frameworks (PAFs) via the cyclotrimerization of methyl ketones mediated by molten ZnCl₂.⁴⁵

Despite being a prevalent method for accessing porous organic materials, ionothermal syntheses have seen limited application for the preparation of MOFs, with a majority of existing reports focused on the preparation of ZIFs and simple carboxylate MOFs, often utilizing ionic liquids as solvent or templating agents.⁴⁶ One report of the preparation of MIL-100(Cr) without the use of solvent or exogenous HF was particularly inspiring to our lab, as the authors report obtaining high-quality MOF material after simply heating CrCl₃‧6H₂O and the corresponding linker together in a Teflon autoclave at temperatures exceeding the melting point of the metal salt.⁴⁷ Encouraged by this finding, we sought to prepare azolate and salicylate MOFs—materials that are especially promising for catalysis but are limited by dilute (< 0.01 M) synthesis conditions and long (> 7 days) reaction times—without solvent, utilizing low-melting molten metal (hydrate) salts as ionothermal media (Figure 5).⁴⁸

Figure 5.

a) Ionothermal synthesis of Fe₂X₂(dobdc) and Fe₂X₂(m-dobdc) (X=Cl, OH). Structures of Fe₂X₂(dobdc)and Fe₂X₂(m-dobdc) (X=Cl, OH). Gray, white, red, orange, and green spheres correspond to carbon, hydrogen, oxygen, iron, and chlorine, respectively. b) SEM images of Fe₂X₂(dobdc) and Fe₂X₂(m-dobdc) (X=Cl, OH) prepared under ionothermal conditions. c) Mössbauer profile of the Fe(III) MOFs Fe₂X₂(m-dobdc) (X=Cl, OH). The Mössbauer profile of the Fe(II) MOF Fe₂(dobdc) is included for comparison. Adapted under terms of the CC-BY license. Copyright 2023, published by Wiley-VCH GmbH, Angewandte Chemie International Edition.⁴⁸

Starting with azolate MOFs, we identified the open-metal site M₂Cl₂(btdd) (M = V, Mn, Fe, Co, Ni, Cu) family of MOFs as promising candidates for ionothermal synthesis.⁸ Indeed, combining the corresponding metal chloride hydrate salts with the H₂btdd linker at temperatures above the melting point of the salt (160 °C), yields Co₂Cl₂(btdd) and Ni₂Cl₂(btdd) after 16 h. Notably, the temperatures at which these reactions are performed is below the melting or decomposition points of the corresponding linkers.⁴⁸ These samples can be purified from soluble impurities via Soxhlet extraction. In addition to their ease of preparation, these samples exhibit very high BET surface areas (Table 3) and match their solvothermal counterparts by PXRD and SEM. This ionothermal approach is also scalable, exemplified by the preparation of upwards of 1.5 g of Co₂Cl₂(btdd) in a single Teflon autoclave. Initial attempts to prepare the Zn polymorph of these MOFs, Zn₅Cl₄(btdd)₃, also known as MFU-4l,⁴⁹ resulted in graphitization and degradation of the framework due to the high melting point of anhydrous ZnCl₂ (290 °C). This decomposition can be avoided by utilizing a 6:1:1 eutectic mixture of ZnCl₂, NaCl and KCl (melting point = 225 °C), resulting in material of comparable quality to traditionally prepared MFU-4l. Attempts to prepare the Mn- and Cu-based analogues were unsuccessful, but Ni-based frameworks constructed from other azolate linkers, including Ni₃(btp)₂ (btp³⁻ = 4,4’,4’’-(benzene-1,3,5-triyl)tris(pyrazolate)⁵⁰ and Ni(bdp) (bdp²⁻ = 1,4-benzene-4,4′-dipyrazolate)⁵¹ could be synthesized. These findings indicate that the ionothermal synthesis of thermodynamically stable azolate frameworks constructed from strong M–N bonds is generalizable.

Table 3.

77 K N₂ BET surface areas of MOFs prepared ionothermally. Experimental values reported for solvothermal equivalents are included for reference.⁴⁸

MOF	BET Surface Area (m2/g)
	Ionothermal	Solvothermal
Co2Cl2(btdd)	2321	2438
Ni2Cl2(btdd)	2438	1973
Zn5Cl4(btdd)3	2524	2750
Ni3(btp)2	2602	2100
Co2(dobdc)	1042	1135
Fe2X2(dobdc)	645	--
Ni2(m-dobdc)	1416	1592
Fe2X2(m-dobdc)	809	--

We then turned our attention to salicylate-based frameworks, chiefly M₂(dobdc)⁵² and M₂(m-dobdc).⁴¹ Initial attempts to synthesize Co₂(dobdc) and Ni₂(m-dobdc) at 160 °C failed; however, these MOFs can be readily obtained by simply increasing the reaction temperature to 200 °C. These MOFs possess BET surface areas comparable to the corresponding MOFs prepared under traditional solvothermal conditions (Table 3) and show a notable lack of residual chloride by X-ray photoelectron spectroscopy (XPS). Attempts to prepare less robust azolate frameworks (Mn₂Cl₂(btdd), Cu₂Cl₂(btdd)) and salicylate frameworks (Mg₂(dobdc), Mn₂(dobdc), Cu₂(dobdc)) were unsuccessful, indicating that acid-stability is paramount when choosing candidate materials to synthesize via ionothermal methods.

Building upon these results, we hypothesized that the unusual kinetic regime offered by ionothermal methods might enable the synthesis of new frameworks that cannot be synthesized directly under solvothermal conditions. Indeed, we reported the first preparations of Fe₂X₂(dobdc) and Fe₂X₂(m-dobdc) (X = Cl, OH), which are Fe(III) analogues of the highly air-sensitive MOFs Fe₂(dobdc) and Fe₂(m-dobdc), under ionothermal conditions (Figure 5a). Attempts to prepare these Fe(III) salicylate MOFs under solvothermal conditions resulted in amorphous solids. These new materials exhibit Mössbauer profiles characteristic of Fe(III) materials (Figure 5c) and BET surface areas that are reasonably attenuated compared to their open-metal-site counterparts due to the presence of anions in their pores (Table 3). The anions are likely a mixture of Cl− and OH−, as suggested by a combination of combustion elemental analysis and energy-dispersive X-ray spectroscopy (EDS). Overall, this ionothermal approach represents a sustainable alternative to traditional solvothermal syntheses that does not require complex instrumentation or strenuous mechanochemical conditions.

6. Example synthesis of Mg₂(dobdc).

Although our primary goal throughout the development of the synthetic methods outlined above was to identify general approaches for scalable MOF synthesis, our efforts were driven in part by the need to make a single MOF—Mg₂(dobdc)—in bulk.⁵³ This lightweight framework bears a high density of coordinatively unsaturated Mg²⁺ sites and thus exhibits high gravimetric and volumetric adsorption capacities for a range of gases. We identified Mg₂(dobdc) as a promising framework for the storage and delivery of fluorinated gases relevant to pharmaceutical synthesis.⁵⁴ As these reactions require stoichiometric amounts of MOF, we needed a way to reliably prepare decagrams of Mg₂(dobdc) in a single batch. Below, we describe our efforts to identify such a method to serve as a road map for others interested in synthesizing MOFs on scale.

Scattered large-scale syntheses of Mg₂(dobdc) with variable surface areas have been reported,^55,56 but it is still often prepared under ultra-dilute solvothermal conditions (~0.01 M in DMF/alcohol mixtures) in a jar without stirring. The latter approach works well to produce small quantities of Mg₂(dobdc) (Langmuir surface area = 1914 m²/g) and can be translated to a round-bottom flask equipped with a reflux condenser to enable the gram-scale synthesis of high-quality Mg₂(dobdc) (Langmuir surface area = 1947 m²/g).⁵⁷ Unfortunately, we found that this approach tends to produce unidentified phases on larger (~5 g) scales, likely due to poor mixing of the liters of solvent required. Extensive soaking in methanol is also needed to completely remove strongly bound DMF from the pores. As such, we aimed to identify a reliable and scalable synthesis of Mg₂(dobdc) that employs smaller volumes of organic solvents, with the ultimate goal of bypassing the use of DMF entirely.

We first turned to mechanochemistry, as several mechanochemical syntheses of Mg₂(dobdc) and related MOFs have been reported.⁵⁸ As outlined in Section 4, we found that amines can serve as both the base required to deprotonate H₄dobdc and the liquid needed for LAG, enabling the rapid synthesis of porous Mg₂(dobdc) and related MOFs via mechanical grinding in a planetary ball mill.^42,39 High surface area Mg₂(dobdc) (Langmuir surface area = 1992 m²/g) can be produced on gram scale in just 5 min of grinding time using this approach. The main limitation of this strategy is the need for specialized equipment: further scaling up of this reaction would have necessitated a ball mill with larger milling jars. As such, we aimed to improve the scalability of the solvothermal synthesis of Mg₂(dobdc), which does not require any equipment beyond a flask and stir bar.

As discussed in Section 3, attempted synthesis of Mg₂(dobdc) at > 0.1 M linker concentrations leads instead to a new framework, CORN-MOF-1 (Mg), that can subsequently be converted to Mg₂(dobdc) by heating in DMF.³² This two-step route reproducibly enables the synthesis of high-quality Mg₂(dobdc) (Langmuir surface area = 2060 m²/g) on ~5 g scale and thus became our preferred approach to prepare this MOF. However, this synthetic route does suffer from two major drawbacks: 1) the use of DMF, which necessitates extensive soaking in methanol, and 2) long reaction times, as at least 5 days of heating in DMF are needed to fully convert CORN-MOF-1 (Mg) to Mg₂(dobdc). In our hands, the volume of methanol required for soaking could not be minimized using Soxhlet extraction due to degradation of the MOF under these conditions.

Building upon these high-concentration results (Section 3) and our mechanochemical studies (Section 4), we hypothesized that the addition of exogenous base at high reaction concentrations should help to bypass the formation of CORN-MOF-1 (Mg), which has partially deprotonated linkers. Indeed, we found that the addition of sodium hydroxide enables the synthesis of Mg₂(dobdc) under high-concentration (1.0 M) conditions in water at room temperature, eliminating the use of DMF entirely. This route was employed to produce > 100 g of Mg₂(dobdc) in a single batch (Langmuir surface area = 2379 m²/g) in just 24 h of synthesis time. Based on this success, we expect that many other MOFs can be prepared under high-concentration conditions, potentially with the addition of stoichiometric base.

7. Conclusion.

As outlined above, many user-friendly avenues to the scalable synthesis of MOFs, including high-concentration solvothermal (Section 3), mechanochemical (Section 4), and ionothermal (Section 5) syntheses have emerged in recent years. Each of these approaches possesses its own advantages and potential drawbacks. If a researcher is starting with a known solvothermal synthesis and aiming to increase the amount of MOF that can be prepared in a single batch, the easiest route to try first is high-concentration solvothermal synthesis (Section 3). In this manner, no other reaction parameters besides the amounts of reagents employed need to be changed to potentially produce MOFs on gram scale. However, as noted in our discussion about the attempted high-concentration synthesis of Mg₂(dobdc), this approach may lead to kinetic phases if the desired phase is not thermodynamically favored enough at high reaction concentrations. Mechanochemical synthesis offers the advantages of fast reaction times and minimal solvent use, but tends to produce poorly crystalline MOFs and requires specialized equipment such as a ball mill or twin-screw extruder, which may be constrictive for researchers without ready access to these instruments. On the other hand, ionothermal MOF synthesis does not require any specialized equipment, but it does necessitate access to low-melting metal salt precursors to enable MOF formation in the molten metal salt medium. Overall, access to multiple avenues to easily produce MOFs on gram scale will guarantee that at least one route will work for a given framework.

Moving forward, more in-depth studies into the chemical pathways that guide self-assembly, such as the role of acid modulators (Section 2), are needed to streamline access to MOFs with ever-improving properties. We encourage all researchers in the field to consider the user-friendliness and potential scalability of all reported MOF syntheses to enable others, especially non-experts, to prepare and study MOFs for next-generation applications.

Biographies

Tyler Azbell is a Ph.D. candidate in Chemistry at Cornell University. Originally from Columbus, Ohio, he received his bachelor’s degree in chemistry from Purdue University in 2019. His research is currently focused on redox-active open-metal site MOFs, catalysis, and sustainability.

Tristan Pitt is a Ph.D. candidate in Chemistry at Cornell University. Originally from Boise, Idaho, he received his bachelor’s degree in chemistry from Carleton College in 2019, then spent one year working at Los Alamos National Laboratory. His research is currently focused on controlling MOF self-assembly, synthesizing covalent organic frameworks, and gas capture.

Ronald Jerozal is a Ph.D. candidate in Chemistry at Cornell University. Originally from Buffalo, New York, he received his B.S. in Chemistry and B.A. in Mathematics from University at Buffalo in 2019. His research is currently focused on sustainable MOF synthesis, catalysis, and gas capture.

Ruth Mandel is a Ph.D. candidate in Chemistry at Cornell University. Originally from Bucks County, Pennsylvania, she received her bachelor’s degree in chemistry from the University of Delaware in 2019. Her research is currently focused on gas storage, delivery, and separations in MOFs.

Phillip Milner received B.A.s in Chemistry and Mathematics from Hamilton College in 2010. He earned his PhD in 2015, working under the guidance of Prof. Stephen Buchwald at MIT. He then completed post-doctoral studies in the laboratory of Prof. Jeffrey Long at the University of California, Berkeley. He has been a professor at Cornell University since 2018, where his research interests lie at the interface of organic, inorganic, and materials chemistry.